Effects of Introduced Mosquitofish and Bullfrogs on the Threatened California Red-Legged Frog

SHARON P. LAWLER,*‡ DEBORAH DRITZ,* TERRY STRANGE,† AND MARCEL HOLYOAK* *Department of Entomology, University of California, Davis, CA 95616–8584, U.S.A. †San Joaquin County Mosquito and Vector Control District, 7759 South Airport Way, Stockton, CA 95206, U.S.A.

Abstract: Exotic species have frequently caused declines of native fauna and may contribute to some cases of amphibian decline. Introductions of mosquitofish (Gambusia affinis) and bullfrogs (Rana catesbeiana) are suspected to have caused the decline of California red-legged frogs (Rana aurora draytonii). We tested the effects of mosquitofish and bullfrog tadpoles on red-legged frog tadpoles in spatially complex, speciose communities. We added 720 hatchling red-legged frog tadpoles to each of 12 earthen ponds. Three ponds were controls, 3 were stocked with 50 bullfrog tadpoles, 3 with 8 adult mosquitofish, and 3 with 50 bullfrogs plus 8 mosquitofish. We performed tests in aquaria to determine whether red-legged frog tadpoles are preferred prey of mosquitofish. Mosquitofish fed on a mixture of equal numbers of tadpoles and either mosquitoes, Daphnia, or corixids until Ͻ 50% of prey were eaten; then we calculated whether there was disproportionate predation on tadpoles. We also recorded the activity of tadpoles in the presence and absence of mosquitofish to test whether mosquitofish interfere with tadpole foraging. Survival of red-legged frogs in the presence of bullfrog tadpoles was less than 5%; survival was 34% in control ponds. Mosquitofish did not affect red-legged frog survival, even though fish became abundant (approximately 1011 per pond). Two mechanisms may have blocked the effects of mosquitofish on tadpole survival: (1) fish ponds contained fewer predatory invertebrates, and (2) mosquitofish preferred other prey to red-legged frogs in laboratory trials. Red-legged frog tadpoles suffered more injuries in ponds with fish, however, and weighed 34% less at metamorphosis. The growth decrease could have been caused by injuries or by lower foraging levels in the presence of fish. Labo- ratory results showed that young tadpoles were less active in the presence of mosquitofish. Although both mosquitofish and bullfrogs affected red-legged frogs, the impact of bullfrogs on the survival of red-legged frogs may contribute more strongly to their decline.

Effectos de la Introducción del Pez Mosquito y Rana Toro en Poblaciones de la Rana Resumen: Especies exóticas han ocasionado frecuentemente disminuciones de fauna nativa y pueden con- tribuir en algunos casos a la disminución de anfibios. Introducciones del pez mosquito (Gambusia affinis) y rana toro ( Rana catesbeiana) son consideradas como la causa del declive de la rana patiroja de California ( Rana aurora draytonii). Evaluamos los efectos del pez mosquito y renacuajos de rana toro en renacuajos de rana patiroja en comunidades espacialmente complejas. Agregamos 720 renacuajos recién eclosionados a cada uno de los 12 estanques de tierra. Tres de los estanques eran controles, tres fueron sembrados con 50 renacuajos de rana toro, tres con 8 adultos de pez mosquito y 3 con 50 renacuajos de rana toro mas ocho peces mosquito. Realizamos pruebas en acuarios para checar si los renacuajos de patiroja eran la presa preferida del pez mosquito. Los peces mosquito se alimentaron de una mezcla de igual número de renacuajos y tanto mosquitos, Daphnias, o coríxidos, evaluamos si hubo una depredación desproporcionada de renacuajos una vez que alrededor del 50% de las presas fué comido. También observamos la actividad de los renacuajos en presencia y ausencia de peces mosquito para determinar si los peces interferían con el forrajeo de los renacuajos. La supervivencia de ranas patiroja en presencia de renacuajos de rana toro fue redu-

‡email [email protected] Paper submitted February 20, 1998; revised manuscript accepted August 26, 1998. 613

Conservation Biology, Pages 613–622 Volume 13, No. 3, June 1999

614 Effects of Alien Species on Red-Legged Frogs Lawler et al. cida en un 5%; la supervivencia en los estanques control fué de un 34%. El pez mosquito no afectó la super- .(por estanque 1011ف) vivencia de la rana patiroja, aún cuando la abundancia de peces se tornó elevada Dos mecanismos pueden haber bloqueado los efectos del pez mosquito en la supervivencia de renacuajos: (1) los estanques con peces tuvieron menos invertebrados depredadores y (2) los peces mosquito prefírieron otro tipo de presas en los experimentos de laboratorio. Sin embargo, los renacuajos de rana patiroja sufrieron mas lesiones en estanques con peces y pesaron 34% menos durante metamorfósis. La disminución en crec- imiento pudo haber sido causada por lesines o por niveles mas bajos de forrajeo en presencia de peces. Los resultados de laboratorio muestran que los renacuajo jóvenes fueron menos activos en presencia de peces mosquito. A pesar de que tanto los peces mosquito como la rana toro afectaron a la rana patiroja, el impacto de la rana toro en la supervivencia de la rana patiroja contribuye mas fuertemente a su disminución.

Introduction In a field survey of sites where red-legged frogs have bred historically, Hayes and Jennings (1986) found a Introduced species have caused declines or extinctions strong negative correlation between extant populations of native species worldwide. Some of the most ecologi- of the frog and the presence of introduced fish and bull- cally destructive introductions have been those in which frogs. We measured how tadpoles of red-legged frogs the alien species is a vertebrate that was introduced for are affected by bullfrogs and mosquitofish (Gambusia economic reasons (Williamson 1996) or when a general- affinis affinis). We selected the mosquitofish because ist predator is introduced as a biological control agent of its widespread and ongoing introduction (Rupp (Howarth 1991; Simberloff & Stiling 1996). Species are 1996); other species of fish may also play a role in the rarely introduced singly or in the absence of other decline of red-legged frogs. changes, such as habitat destruction, and this can make Bullfrogs are native to the eastern United States. They it difficult to relate the decline of a native species to the were introduced in the west to improve the frog “fishery” introduction of a particular alien species (Williamson and have been introduced to Europe as well (Stumpel 1996). Experimental work is necessary to estimate the 1992). Bullfrog tadpoles can be strong competitors effects of any one factor, preferably by manipulating one (Werner 1994; Kupferberg 1995). Red-legged frog tad- or more species independently under conditions that poles face competition from large, overwintering bull- are as natural as possible. We present a field experiment frog tadpoles because bullfrogs breed in summer and that quantified the effects of two introduced vertebrate red-legged frogs breed from January to March. Larger species on tadpoles of an endangered frog. tadpoles sometimes enjoy a greater per-capita competi- Non-native vertebrate species are suspected to have tive advantage (Lawler & Morin 1993; Werner 1994; contributed to the decline of the California red-legged Kupferberg 1995; Peacor & Werner 1997). In addition, frog (Rana aurora draytonii), which is listed as a threat- bullfrog tadpoles may consume northern red-legged frog ened species. The decline of this formerly abundant spe- tadpoles (Rana aurora aurora) under some circum- cies is one of the most dramatic examples of amphibian stances (Kiesecker & Blaustein 1997). The effects of decline (A. R. Blaustein et al. 1994), and examining causes bullfrog adults on native frogs may be even more pro- for this phenomenon could provide clues that may to aid nounced because bullfrogs will consume other frogs. Ex- in the recovery of other declining amphibians. Histori- perimental studies are needed to determine the effect of cally, R. a. draytonii bred throughout the lower central bullfrogs on other amphibians because several studies valley of California and in low elevations in the Sierra and have documented the decline of native species after coastal ranges, but it is now restricted largely to the bullfrog introductions (Kiesecker & Blaustein 1997). coastal foothills ( Jennings & Hayes 1985; Fisher & Shaffer Mosquitofish are native to the eastern United States, 1996). A dramatic early decline began in 1895 and was as- and have been introduced to wetlands worldwide as bio- sociated with human consumption of frogs, but the con- logical control agents for mosquito larvae. The practice of tinued loss of populations is poorly understood ( Jennings stocking mosquitofish concerns conservationists because & Hayes 1985). Possible mechanisms include habitat loss introduced mosquitofish can harm amphibians (e.g., or degradation, competition with and predation from the Woodward 1983; Gamradt & Kats 1996) and other native bullfrog (Rana catesbeiana), and predation by intro- species (Courtenay & Meffe 1989; Rupp 1996). Mosqui- duced fishes (Hayes & Jennings 1986). Some researchers tofish are capable of eliminating red-legged frog tadpoles have concluded that introduced predators are the most from simple communities in small artificial pools even likely cause of red-legged frog disappearance, although when tadpoles are as large or larger than the mosqui- multiple factors probably contribute to the decline (Hayes tofish (R. Schmeider & R. Nauman, personal communi- & Jennings 1986; Fisher & Shaffer 1996). cation). Mosquitofish can also injure or kill fish larger

Conservation Biology Volume 13, No. 3, June 1999

Lawler et al. Effects of Alien Species on Red-Legged Frogs 615 than themselves (Courtenay & Meffe 1989), so it is unlikely Because red-legged frogs typically breed in areas with that tadpoles could outgrow mosquitofish predation. emergent vegetation, we planted 20 cattails along the The evidence that mosquitofish may play a role in the southeast corner of each pond and added 100 old cattail decline of the red-legged frog is inconclusive because it is leaves to the ponds. An irrigation system with float-valves based on correlations between species distributions and kept the ponds full of well water. We added a 1-L sample on predation trials in arenas where habitat and commu- of zooplankton and algae from a nearby wetland to all nity structure were simplified. Habitat structure can be ponds (concentrated from 20, 2-m plankton net tows), important to prey persistence (e.g., Crowder & Cooper and aquatic insects colonized ponds naturally. We in- 1982; Nelson & Bonsdorff 1990; Holyoak & Lawler 1996), stalled aluminum fences 60 cm high around ponds to pre- and biotic complexity can allow indirect effects that vent juvenile frogs from emigrating. make it difficult to predict whether an observed direct ef- On 15 March 1996 we added 720 newly hatched red- fect of a predator on a prey will translate into altered pop- legged frog tadpoles to each pond, which is within the ulation dynamics (Wootton 1994; Menge 1995; Polis & range of natural densities for this species. Hatchlings Strong 1996). The mosquitofish is a generalist predator were obtained from eggs collected in San Joaquin and that consumes a wide variety of prey (Bence 1988; Linden Contra Costa Counties. We used hatchlings because red- & Cech 1990; Rupp 1996), including predators that eat legged frogs lay their eggs in firmly cohesive bunches tadpoles (e.g., odonates; Smith 1983; Travis et al. 1985) that would be difficult to separate and count without and invertebrates that compete with tadpoles (e.g., mos- damaging the developing eggs. Mosquitofish are usually quitoes; Morin et al. 1988; L. Blaustein & Margalit 1994). unable to remove embryos from large ranid eggs like The presence of alternative prey might either reduce the red-legged frog eggs (Grubb 1972), but using hatchlings impact of the predator on a particular prey (Murdoch could produce an underestimate of predation if mosqui- 1969) or increase predation if additional prey support a tofish feed at hatching egg masses. Hatchlings from six higher abundance of the predator (Holt 1977; Lawler clutches were mixed before they were counted and 1993; Holt & Lawton 1994). Predators can sometimes added to ponds to minimize variation among replicates shift the competitive balance between amphibians be- due to genetic or parental effects. On the same day, cause tadpoles of some species forage less and grow more three ponds also received 50 large bullfrog tadpoles slowly in the presence of predators (Woodward 1983; (mean total length 10.98 cm Ϯ 1.25 SD), three received Lawler 1989; Werner 1991). eight adult mosquitofish (four males and four pregnant The crucial question is whether the sum of the direct females), three received both bullfrogs and fish, and and indirect effects of the introduced species on red- three controls received no additional vertebrates. Over- legged frogs is positive or negative. To address this ques- wintering bullfrog tadpoles were collected from a pond tion we raised red-legged frog tadpoles in replicated, in Stockton, California. Metamorphosing bullfrogs were spatially complex ponds that contained the types of al- collected between days 75 and 256. Mean day of meta- ternative prey and other predators commonly found in morphosis was 107 Ϯ 26 SD. A mean of 26 Ϯ 5 SD bull- red-legged frog breeding sites, and we compared the frogs completed metamorphosis in the ponds, and fish performance of red-legged frog tadpoles in these ponds did not affect the number, weight, or date of the meta- and in similar ponds with mosquitofish, or bullfrogs, or morphosis of bullfrogs. both. We also collected information on numbers of in- The initial number of mosquitofish exceeded the rec- vertebrate predators and competitors in experimental ommended stocking rate of 0.045 kg/acre used by mos- ponds to assess the likelihood of various indirect effects. quito abatement districts, but it is in keeping with their To further aid our interpretation of the results of the practice of introducing 6–10 fish to small bodies of wa- pond experiment, we conducted laboratory tests on the ter such as cattle watering tanks to ensure that a breed- feeding preferences of mosquitofish and the behavior of ing population will result. Mosquitofish populations are tadpoles. often low early in the year because of natural winter mortality and flushing of fish from pools during winter floods (Meffe 1983; Swanson et al. 1996). Schools of Methods adult and young mosquitofish were visible in all fish ponds within 3 weeks of introduction. On day 124, three observers counted the number of fish Ͼ1 cm in to- Pond Study tal length visible in 25% of the surface of ponds within We constructed 12 earthen ponds in San Joaquin County, the top 5 cm. We counted an average of 51 Ϯ 8 SE mos- California, during the fall of 1995. Each pond was 3.05 ϫ quitofish per pond. Ponds therefore contained an average 6.10 m and sloped gradually to 1.22 m deep at one end. of 200 fish by that date. This is probably an underestimate Pond bottoms were sealed with clay. We covered approx- because mosquitofish also use deeper waters. We esti- imately 2 m2 of the substrate in the shallow end of ponds mated final densities in November (day 249) via mark- with cobblestones approximately 10–25 cm in diameter. release-recapture. We introduced 100 albino mosquitofish

Conservation Biology Volume 13, No. 3, June 1999

616 Effects of Alien Species on Red-Legged Frogs Lawler et al. to each pond as marked fish, allowed them to mingle by the variable weights of frogs from the mosquitofish- with the normal fish for 1 hour, and then recaptured plus-bullfrog treatment. We therefore compared the Ն200 fish to determine the proportion marked. To as- weights of red-legged frogs from controls to those from sess the accuracy of the study we performed a second the bullfrog-only and mosquitofish-only treatments sepa- mark-recapture in three ponds in which we clipped a rately and discarded the bullfrog-plus-mosquitofish data. small section of the tail fin to mark 100 resident fish. We This gave us more power to detect a fish effect. We used drained these three ponds 1 week later and counted all a repeated-measures ANOVA to determine whether fish fish. The study using albinos showed a mean of 1011 Ϯ affected the date of metamorphosis of red-legged frogs, 139 SE mosquitofish per pond. Analysis of variance again omitting data from ponds with bullfrogs because (ANOVA) showed that the final census did not differ sig- few frogs emerged from these ponds and the dates of nificantly from the albino study (ANOVA F ϭ 1.54, df ϭ emergence were quite variable. Because we analyzed 1,4, p Ͼ 0.28), but the fin-clip method showed a ten- three response variables for the red-legged frogs in this dency to underestimate abundance compared to the experiment, we applied a sequential Bonferroni correc- census (ANOVA F ϭ 5.27, df ϭ 1,4, p Ͻ 0.09). Bullfrog tion to the alpha levels (Sokal & Rohlf 1995); these are tadpoles had no detectable effect on fish abundance given in the statistical tables. (ANOVA F ϭ 0.45, df 1,4, p ϭ 0.535). We collected sweep-net samples of invertebrates from Laboratory Experiments all ponds on days 20, 35, 49, 63, 77, 116, and 144. For each sample we took two benthic sweeps and two midwa- Mosquitofish feeding-preference experiments were con- ter sweeps of approximately 2 m, with a standard d-ring ducted in 20-L aquaria containing spring water, a layer of net (1-mm mesh). These data were analyzed with repeated- gravel 2 cm deep, and a plastic plant. Three sides were measures ANOVA (SYSTAT, 1992). To check whether covered with paper, and the fourth was covered with mosquitofish differentially affected invertebrates that might one-way mirror film. Tanks were lit from above with a eat or compete with red-legged frog tadpoles, we divided 15-watt fluorescent lamp. Each aquarium was divided the data set into two categories for separate repeated- with a removable opaque barrier, which allowed fish measures ANOVA: “dominant predatory invertebrates” and their prey to acclimate in different compartments (odonates, belostomatids, coleopterans, and notonectids) before trials. The barriers were not watertight. Red- and “dominant grazing invertebrates” (ephemeropterans, legged frog tadpoles were obtained from eggs found in a corixids, chironomids, and gastropods). We further quanti- San Joaquin County pond and held in predator-free fied differences in the abundance of predatory insects on ponds. Mosquitofish were provided by local mosquito 17 July by collecting dragonfly exuviae (cast skins) from abatement districts and were fed flake food. All fish pond vegetation and by counting backswimmers visible in were nonpregnant, mature females, and they were not 25% of the pond surface. fed for 24 hours preceding trials. Individual organisms During the study, we noticed tail injuries in red-legged were used only once in experiments. frog tadpoles. On day 160, we collected a sample of 17 For each trial, an individual fish was placed in an tadpoles from each of two ponds with mosquitofish and aquarium behind the opaque divider. Fifteen tadpoles 6 tadpoles from each of two without mosquitofish to as- and 15 alternative prey were added to the other side of sess them for tail damage. Bullfrogs were not present in the divider and were allowed to acclimate for 15 min- these ponds. Sample sizes differed because fewer tad- utes. We then removed the divider and observed fish poles remained in ponds without mosquitofish. feeding behavior for up to 30 minutes or until 50% of Red-legged frogs and bullfrogs metamorphosed into the prey had been consumed. There were five or six tri- juveniles between June and November 1996. We col- als for each species of alternative prey. For all feeding- lected the juveniles on 1–3 nights per week. Juveniles preference trials, we analyzed the data using replicated were counted and weighed after they resorbed their goodness-of-fit tests (Sokal & Rohlf 1995). tails. Red-legged frogs were fed and released in newly es- We compared attacks by mosquitofish on red-legged tablished mitigation ponds in Contra Costa County. Bull- frog hatchlings versus mosquito larvae (Culex tarsalis) frogs were donated to another research project. or water boatmen (Corixidae). Hatchling tadpoles mea- We used SYSTAT statistical software (SYSTAT Inc. sured 1–1.5 cm total length, mosquito larvae were ap- 1992) to perform ANOVA on red-legged frog survival proximately 0.9 cm long, and corixids were approxi- and weight data. The proportion surviving in each pond mately 0.7 cm long. We repeated the mosquito trial was arcsin-square-root-transformed before a fully-facto- using larger, more active tadpoles that averaged 4.3 cm rial ANOVA. Weights of juvenile frogs were ln-trans- total length, developmental stage 26 (Gosner 1960), and formed before analysis. A fully-factorial ANOVA on red- included an additional trial using Daphnia magna (0.2– legged frog weights did not detect the effects of bull- 0.3 cm long) as the alternative prey. Tadpoles used in frogs or mosquitofish, but inspection of the data indi- these trials were too large for the fish to engulf, but they cated that an effect of mosquitofish had been obscured were still vulnerable to attack.

Conservation Biology Volume 13, No. 3, June 1999

Lawler et al. Effects of Alien Species on Red-Legged Frogs 617

A related experiment conducted over a 16-hour pe- Table 1. ANOVA of the effects of bullfrog tadpoles and riod tested whether the presence of alternative prey mosquitofish on the proportion of red-legged frog tadpoles that survived to become juvenile frogs.* would affect mosquitofish consumption of tadpoles. We placed three fish and three stage-26 tadpoles (similar in Mean size to those of the first feeding-preference trials) in Source df square Fp each of eight aquaria. Four aquaria also received 40 late Bullfrogs 1 0.75 88.18 Ͻ0.001 instar Culex larvae. We used several fish together in this Mosquitofish 1 0.00 0.01 0.898 experiment because mosquitofish often forage in small Bullfrogs ϫ mosquitofish 1 0.00 0.79 0.399 groups. The experiment lasted from 1800 hours until Error 8 0.00 1000 hours the next day, at which time prey were *Data were arcsin-square-root-transformed before analysis. The se- counted and checked for injury. Data were analyzed quential Bonferroni-corrected alpha level for this test is 0.016. with the general linear model program GLIM (Baker 1987). This program allowed us to assume a binomial to emerge from ponds with bullfrogs approximately 20 distribution of sampling error and to scale the model by days after they first emerged from the other ponds (Fig. mean square error so that changes in deviance were as- 1). Although the dates of first emergence were compara- sumed to follow a chi-square distribution. We performed ble in ponds with red-legged frogs alone and in those an ANOVA on logit-transformed proportions of tadpoles with mosquitofish, a repeated-measures ANOVA showed with tail damage. that emergence lagged in ponds with mosquitofish after To determine whether mosquitofish affect tadpole for- the first few weeks (Table 2; Fig. 1). aging, we compared the activity level of tadpoles in the There was no detectable effect of bullfrog tadpoles on presence and absence of fish in two experiments, one the weight of juvenile red-legged frogs (Table 3). The when tadpoles were at developmental stage 26 and weights of the few red-legged frogs that emerged from mean length 2.4 cm and the second when tadpoles were ponds with bullfrogs were variable. In contrast, the at developmental stage 33–36 and at a mean length of frogs that emerged from ponds with mosquitofish 5.9 cm. The stage-26 tadpoles in this trial were much weighed 34% less than those emerging from control smaller than those used in the second feeding trial de- ponds (Table 3; Fig. 2). scribed above, but it is not unusual to find a large range At least part of the weight decrease in frogs raised of body sizes in the early developmental stages because with fish might be explained by injuries inflicted on the tadpoles accomplish much of their growth during these tadpoles by mosquitofish. A sample of tadpoles col- stages. For each trial, a single tadpole and either nothing (control) or a single fish were added to different compartments of an aquarium. Animals acclimated over- night. The next day, the investigator removed the divider and recorded tadpole activity every 15 seconds for 10 minutes. There were eight control trials and eight fish trials for each experiment. Data were analyzed via GLIM with a logit link function of the proportion of observations in which tadpoles were active. Maximum like- lihoods were calculated by a fitting procedure in which changes in deviance were assumed to follow a chi- square distribution.

Results

Pond Study Bullfrog tadpoles decreased the survival of red-legged frogs. Fewer than 5% of red-legged frogs survived in Figure 1. Mean cumulative survival to metamorpho- ponds with bullfrog tadpoles, in contrast to 30–40% sur- sis of 720 red-legged frog tadpoles raised in ponds vival in other ponds (Table 1; Fig. 1). Mosquitofish had alone, with mosquitofish, with bullfrog tadpoles, and no detectable effect on red-legged frog survival, and we with both mosquitofish and bullfrog tadpoles (r, could not detect any interaction between the effects of ponds with red-legged frogs alone; mr, ponds with bullfrogs and mosquitofish on red-legged frog survival. mosquitofish; br, ponds with bullfrogs; and mbr, The presence of bullfrogs and mosquitofish delayed ponds with mosquitofish and bullfrogs). Error bars red-legged frog metamorphosis. Red-legged frogs began are standard deviations.

Conservation Biology Volume 13, No. 3, June 1999

618 Effects of Alien Species on Red-Legged Frogs Lawler et al.

Table 2. Results of a repeated-measures ANOVA on the effects of mosquitofish (fish) on the number of red-legged frogs emerging from ponds on 39 collection dates.*

Mean Source df square Fp Between subjects Fish 1 88.61 0.65 0.463 Within subjects Day 38 86.69 2.42 0.001 Day ϫ fish 38 56.01 1.56 0.03 Error 152 35.69 *Although the treatment did not affect the number emerging (between-subjects analysis), juveniles emerged at different rates in ponds with versus those without mosquitofish (within subjects, day ϫ fish interaction; also see Fig. 1). The sequential Bonferroni-corrected alpha level for this test is 0.05. lected on day 160 showed that most tadpoles suffered Figure 2. Mean weights of surviving red-legged frog tail damage in ponds with fish. None of 12 tadpoles from tadpoles raised in ponds alone, with mosquitofish, two fishless ponds showed evidence of injury, but 33 of with bullfrog tadpoles, and with both mosquitofish 34 were missing parts of their tails in a sample of tad- and bullfrog tadpoles. Error bars are standard devia- poles from two ponds with fish. tions. The dominant macroinvertebrates found in ponds repre- sented the orders Odonata (Anisoptera and Zygoptera), Hemiptera (Notonectidae, Corixidae, Belostomatidae, Velli- idae and Gerridae), Diptera (Chironomidae, Culicidae), of notonectids and our collection of odonate exuvia on Coleoptera (Dytiscidae, Hydrophilidae), Ephemeroptera 17 July. We were not able to see any notonectids in (Baetidae), and Gastropoda. Invertebrate densities were ponds with mosquitofish, whereas we observed at least lower in ponds with mosquitofish (repeated-measures five notonectids in 25% of the pond surface in all ponds ANOVA on the total numbers of invertebrates in samples without mosquitofish. Exuvia of the dragonfly Anax from ponds with versus without fish: F ϭ 6.20, df ϭ 1,10, were more abundant in ponds without mosquitofish p Ͻ 0.05). Invertebrate abundances increased more over (ANOVA: F ϭ 12.48, df ϭ 1, 10, p Ͻ 0.005). Anax is a time in ponds without fish than in ponds with fish (day-by- large predator; mean exuvium length in our ponds was treatment interaction: F ϭ 2.99, df ϭ 1,10, p Ͻ 0.008). 4.5 Ϯ 0.4 cm SD. Mosquitofish reduced the abundances of predatory invertebrates (Fig. 3 top panel, repeated-measures ANOVA, Laboratory Experiments treatment effect: F ϭ 75.20, df 1,10, p Ͻ 0.001; this differ- ence is significant with a Bonferroni-corrected alpha of Prey-choice trials indicated that the presence of alterna- 0.025). We were unable to detect an effect of mosquitofish tive prey might afford some protection to tadpoles in on grazing invertebrates (repeated-measures ANOVA, natural ponds. Mosquitofish consumed mosquitoes and day-by-treatment interaction: F ϭ 1.37, df ϭ 7, 70, p ϭ corixids in preference to hatchling red-legged frogs (Fig. 0.229; treatment effect: F ϭ 1.04, df ϭ 1, 10, p ϭ 0.331). 4; Culex versus tadpoles, Gp ϭ 77, df ϭ 1, p Ͻ 0.001; The negative effects of mosquitofish on predatory in- corixids versus tadpoles, Gp ϭ 12, df ϭ 1, p Ͻ 0.001). vertebrates were further confirmed by our observations Mosquitofish attacked corixids in preference to tadpoles (Gp ϭ 19.7, df ϭ 1, p Ͻ 0.001) but launched similar numbers of attacks against mosquitoes and tadpoles (G ϭ Table 3. Results of two single-factor ANOVAs on the effects of p bullfrogs and mosquitofish on the weight of juvenile red-legged 0.06, df ϭ 1, not significant). Mosquitofish consumed frogs at metamorphosis.* both Culex and Daphnia in preference to 3-week-old tadpoles (Fig. 4; replicated G test for Culex versus tad- Mean poles, G ϭ 108.1, df ϭ 1, p Ͻ 0.001; Daphnia versus Source df square Fp p tadpoles, Gp ϭ 204, df ϭ 1, p Ͻ 0.001). All attacks di- Bullfrogs 1 0.04 3.56 0.132 rected at Culex or Daphnia resulted in the prey being Error 4 0.01 eaten. In contrast, the larger tadpoles were never con- Mosquitofish 1 0.30 22.03 0.009 Error 4 0.01 sumed and were attacked less frequently than alternative prey (attacks on tadpoles versus Culex, G ϭ 24.5, df ϭ 1, *Data were ln-transformed before analysis. The sequential Bonfer- p roni-corrected alpha level for these tests are 0.05 (bullfrogs) and p Ͻ 0.001; Daphnia versus tadpoles, Gp ϭ 24.1, df ϭ 1, 0.025 (mosquitofish). p Ͻ 0.001). Goodness-of-fit tests for heterogeneity

Conservation Biology Volume 13, No. 3, June 1999 Lawler et al. Effects of Alien Species on Red-Legged Frogs 619

Discussion

Our study provides experimental evidence that bullfrogs may play a role in the decline of the California red- legged frog. The presence of just 50 bullfrog tadpoles nearly precluded recruitment of red-legged frog tadpoles to the juvenile stage in our ponds. We did not identify the mechanism underlying this effect, but strong competitive effects of bullfrog tadpoles are often reported for other species. Alternative mechanisms include predation and parasitism or disease. Bullfrog tadpoles will consume tadpoles of other ranids (Ehrlich 1979), but a study by Kiesecker and Blaustein (1997) showed that bullfrog tadpoles consumed northern red-legged frog tadpoles only if the tadpoles converged on another source of food. The study was conducted in aquaria, so it is unclear how often predation occurs in larger habitats. We cannot rule out the possibility that bullfrog tadpoles may have transmitted an infectious agent to the red-legged frog tadpoles, but bullfrog tadpoles appeared healthy when transferred into our ponds, and we did not observe obviously diseased red-legged frog tadpoles in sweep-net samples. Kupferberg (1995) demonstrated experimentally that shared pathogens were unlikely to be the cause of the negative effects of bullfrogs on foot- hill yellow-legged frogs (Rana boylii). The severe effect of bullfrog tadpoles on red-legged Figure 3. Abundances of invertebrates over time in frog recruitment is probably an underestimate of the to- samples from experimental ponds. Each point repre- tal effect of introduced bullfrog populations on red- sents the average abundance of insects in dipnet sam- legged frogs because adult bullfrogs eat tadpoles and ples of three ponds (Alone, red-legged frogs were the smaller frogs, including other ranid species (Kupferberg only vertebrates; Bullfrog, 50 bullfrog tadpoles were 1995). It is premature to suggest that bullfrogs are the also present; Fish, 8 mosquitofish were also present; most important agent of red-legged frog decline because and Fish ϩ bullfrog, all three vertebrates were the effects of other introduced species have not been present). quantified, and it is also possible that contaminants or diseases play a role. Nevertheless it was sobering to dis- cover that bullfrog tadpoles alone were capable of among replicates showed that tadpoles were nonpre- nearly eliminating recruitment of the native frog in this ferred prey for all but one fish in both experiments. A experiment. Management strategies aimed at removing single fish launched an almost equal number of attacks adult and juvenile bullfrogs may not be effective if even on Culex and tadpoles. a few adults escape and breed; egg masses and tadpoles Fish did not kill stage-26 tadpoles when allowed to for- must be removed as well. age in groups of three during a 16-hour period, but 92% Although mosquitofish can eradicate red-legged frog of tadpoles were injured in tanks without alternative tadpoles from simplified aquatic communities, our ex- prey and 33% of tadpoles were injured in tanks to which periment showed that mosquitofish do not affect the re- Culex had been added. A linear model showed that inju- cruitment of red-legged frogs from more naturalistic, spa- ries were more frequent in tanks without alternative tially complex, and speciose communities in earthen prey ( p Ͻ 0.05). ponds. Nearly equal numbers of juvenile red-legged frogs The presence of fish caused tadpoles to become less emerged from ponds with and without fish, despite dense active early in their development, but this effect disap- mosquitofish populations in the former and injuries to the peared at later stages (Fig. 5). Stage-26 tadpoles were tadpoles that indicated frequent mosquitofish attacks. It is 75% less active in the presence of fish than in controls unlikely that tadpoles outgrew predation because mosqui- (␹2 ϭ 16.74, df ϭ 1, p Ͻ 0.001; treatment explained 54% tofish can kill any size of red-legged frog tadpole (R. of the variance). There was no evidence that larger, Schmeider & R. Nauman, personal communication). Sev- stage 33–36 tadpoles altered their activity level when eral factors could have contributed to this result. First, fish were present (␹2 ϭ 0.65, not significant). tadpoles are not preferred prey of mosquitofish based on

Conservation Biology Volume 13, No. 3, June 1999 620 Effects of Alien Species on Red-Legged Frogs Lawler et al.

Figure 4. Feeding preference of mosquitofish for red-legged frog tadpoles versus invertebrates. Bars show the average number of predation events (consumption or attacks) per fish by five or six mosquitofish that were allowed to inter- act with prey for up to 30 minutes (A and B, stage-26 tadpoles; C and D, hatchling tadpoles). Error bars are Ϯ1 SE. laboratory trials, and alternative prey may have protected tadpoles from lethal levels of mosquitofish attacks. Sec- ond, ponds contained deep areas, cobbles, and dense vegetation that may have provided refuges for the tadpoles (e.g., Peterson et al. 1992; Babbitt & Jordan 1996). Such spatial heterogeneity is present in many areas where red- legged frogs breed, but our experiment may not predict the effects of mosquitofish in shallow, relatively structure- less pools. Third predaceous invertebrates were less abundant in ponds with fish, and any mortality inflicted by the fish might have been balanced by a decrease in predation from other sources. Large, predatory dragonflies (Anax sp.) were common in ponds without fish. Dragonflies can reduce tadpole populations (Heyer et al. 1975; Smith 1983). Caldwell et al. (1980) found that Anax naiads were able to consume tadpoles that were up to 4.9 cm long, which was the maximum size used in their trials. The tadpoles in our Figure 5. Activity level of red-legged frog tadpoles in the study would probably have been vulnerable to dragonfly presence and absence of mosquitofish (A, stage-26 tad- predation because large Anax exuvia were present in poles; B, stage 33–36 tadpoles). Error bars are Ϯ1 SE.

Conservation Biology Volume 13, No. 3, June 1999 Lawler et al. Effects of Alien Species on Red-Legged Frogs 621

June and continued to appear throughout the summer. include naturalistic levels of spatial and biological com- Our ponds also held other invertebrate predators, in- plexity. Although we did not do the many detailed ex- cluding backswimmers, beetles and beetle larvae, belos- periments necessary to prove that indirect effects or spa- tomatids (giant water bugs), and damselfly naiads. tial heterogeneity ameliorate the effect of mosquitofish We may have underestimated mosquitofish effects if on red-legged frog tadpoles, our results differ markedly mosquitofish are initially more abundant in natural from those of predation trials conducted in simplified ponds when tadpoles are small, especially if fish deplete communities. In addition, our observations and labora- alternative prey prior to frog breeding. Our observations tory results demonstrate mechanisms whereby biologi- at two riparian sites in Contra Costa and San Joaquin cal complexity could have changed the effect of mosqui- Counties which have both mosquitofish and red-legged tofish on red-legged frog tadpoles. frogs suggest that mosquitofish densities in early spring Although simplified tests may measure artifacts, we are as low or lower than those in our ponds; these sites support their use to screen predators that may be intro- are in creeks, however, and some mosquitofish may be duced to new areas as biological control agents. Intro- flushed out of creeks by winter rains. duced predators can be so damaging that a “precaution- Although mosquitofish did not affect amphibian sur- ary principle” that errs on the side of protecting native vival in our study, other work by Gamradt and Kats species is best, except in rare cases in which the ecologi- (1996) showed 87% mortality of California newt larvae cal costs of not trying biological control are even greater (Taricha torosa) in stream cages where three mosqui- (Howarth 1991; Simberloff & Stiling 1996; Samways tofish were caged with four newt larvae for 24 hours (n ϭ 1997). Laboratory tests often provide a valuable worst- 6 replicates; cages were 72 ϫ 45 ϫ 21 cm boxes with case estimate of effect (although this is not guaranteed natural substrate). The abundance of alternative prey because predator behavior may be abnormal in an artifi- was not given, and it is unknown whether cage installa- cial environment). However, in cases in which several tion affected alternative prey. These results show that putative causes of a species decline are already in opera- more work is needed to determine the effects of mosqui- tion, tests of single causes in artificial settings may tofish on native fauna. generate red herrings rather than species recoveries. Although mosquitofish did not affect red-legged frog survival, juveniles emerging from ponds with fish metamorphosed later and weighed an average of one-third Acknowledgments less than those raised without fish. Several factors may have contributed to the lower growth rate of tadpoles in We thank B. Eldridge for facilitating this study and J. ponds with mosquitofish. The laboratory trials showed Stroh of the San Joaquin Mosquito and Vector Control that young tadpoles were less active in the presence of District for providing space and aid for the project. The fish. This could have caused a decrease in their initial University of California Mosquito Research Program pro- growth rate (e.g., Werner 1991; Skelly 1992). Evidence vided financial support. We appreciate the hard work of for this mechanism is equivocal because some species of E. Arias, B. Chan, T. Jensen, J. May, B. Peters, C. Vrede- tadpoles also respond to invertebrate predators with de- voe, and “the Lodi Crew” of the San Joaquin M.V.C.D., creased activity (e.g., Lawler 1989; Werner 1991), and who provided technical assistance during the study. We invertebrate predators were common in ponds without thank C. Miller of the Contra Costa M.V.C.D. for contrib- fish. Injuries can also decrease the growth of tadpoles uting the albino mosquitofish. S. Meyers and J. Alvarez, (Wilbur & Semlitsch 1990; Parichy & Kaplan 1992), and of Jones and Stokes Co., and the Los Vaqueros Project fa- injuries were much more common in ponds with fish. cilitated our access to field sites. Several reviewers pro- More work is needed to determine whether mosqui- vided useful comments on the manuscript. tofish pose a threat to red-legged frog populations. The smaller metamorphs emerging from ponds with fish Literature Cited might mature later and lay fewer eggs, as has been demonstrated in other amphibians (Smith 1987; Semlitsch et Babbitt, K. J., and F. Jordan. 1996. Predation on Bufo terrestris tadpoles: al. 1988). We do not know whether the frogs can grow effects of cover and predator identity. Copeia 1996:485–488. quickly enough in the terrestrial environment to com- Baker, R. J. 1987. GLIM 3.77 reference manual. 2nd edition. Numerical Algorithms Group, Oxford, United Kingdom. pensate for their initially smaller size. Because the mos- Bence, J. R. 1988. Indirect effects and biological control of mosquitoes quitofish did have a negative—albeit sublethal—effect on by mosquitofish. Journal of Applied Ecology 25:505–521. red-legged frogs, it is advisable for mosquito control dis- Blaustein, A. R., D. B. Wake, and W. P. Sousa. 1994. Amphibian de- tricts to use other mosquito-control methods in amphib- clines: judging stability, persistence, and susceptibility of popula- ian habitat and surrounding watersheds (e.g., Bacillus tions to local and global extinctions. Conservation Biology 8:60-71. Blaustein, L., and J. Margalit. 1994. Mosquito larvae (Culiseta longiare- thuringiensis israelensis, Lagenidium, B. sphaericus). olata) prey upon and compete with toad tadpoles (Bufo viridis). Our results illustrate that predation experiments de- Journal of Animal Ecology 63:841–850. signed to assess the effects of introduced species should Caldwell, J. P., J. H. Thorp, and T. O. Jervey. 1980. Predator-prey rela-

Conservation Biology Volume 13, No. 3, June 1999 622 Effects of Alien Species on Red-Legged Frogs Lawler et al.

tionships among larval dragonflies, salamanders, and frogs. Oecolo- tween aquatic insects and vertebrates: interaction strength and gia 46:285–289. higher order interactions. Ecology 69:1401–1409. Courtenay, W. R., Jr., and G. K. Meffe. 1989. Small fishes in strange places: Murdoch, W. W. 1969. Switching in general predators: experiments a review of introduced poeciliids. Pages 301–331 in G. K. Meffe and on predator specificity and stability of prey populations. Ecological F. F. Senlson Jr., editors. Ecology and evolution of livebearing fishes Monographs 39:335–354. (Poeciliidae). Prentice Hall, Englewood Cliffs, New Jersey. Nelson, W. G., and E. Bonsdorff. 1990. Fish predation and habitat com- Crowder, L. A., and W. E. Cooper. 1982. Habitat structural complexity plexity: are complexity thresholds real? Journal of Experimental and the interaction between bluegills and their prey. Ecology 63: Marine Biology and Ecology 141:183–192. 1802–1813. Parichy, D. M., and R. H. Kaplan. 1992. Developmental consequences Ehrlich, D. 1979. Predation by bullfrog tadpoles (Rana catesbeiana) of tail injury on larvae of the oriental fire-bellied toad, Bombina on eggs and newly hatched larvae of the plains leopard frog (Rana orientalis. Copeia 1992:129–137. blairi). Bulletin of the Maryland Herpetological Society 15:25–26. Peacor, S. D., and E. E. Werner. 1997. Trait-mediated indirect interac- Fisher, R. N., and H. B. Shaffer. 1996. The decline of amphibians in Cal- tions in a simple aquatic food web. Ecology 78:1146–1156. ifornia’s Great Central Valley. Conservation Biology 10:1387–1397. Peterson, A. G., M. Bull, and L. M. Wheeler. 1992. Habitat choice and Gamradt, S. C., and L. B. Kats. 1996. Effect of introduced crayfish and predator avoidance in tadpoles. Journal of Herpetology 26:142–146. mosquitofish on California newts. Conservation Biology 10:1155– Polis, G. A., and D. R. Strong. 1996. Food web complexity and commu- 1162. nity dynamics. American Naturalist 147:813–846. Gosner, K. L. 1960. A simplified table for staging anuran embryos and Rupp, H. R. 1996. Adverse assessments of Gambusia affinis: an alter- larvae with notes on identification. Herpetologica 16:183–190. native view for mosquito control practitioners. Journal of the Grubb, J. C. 1972. Differential predation by Gambusia affinis on the American Mosquito Control Association 12:155–166. eggs of seven species of anuran amphibians. American Midland Samways, M. J. 1997. Classical biological control and biodiversity con- Naturalist 88:102–108. servation: what risks are we prepared to accept? Biodiversity and Hayes, M. P., and M. R. Jennings. 1986. Decline of ranid frog species in Conservation 6:1309–1316. western North America: are bullfrogs (Rana catesbeiana) respon- Semlitsch, R. D., D. E. Scott, and J. Pechman. 1988. Time and size at sible? Journal of Herpetology 20:490–509. metamorphosis related to adult fitness in Ambystoma talpoideum. Heyer, W. R., R. W. McDiarmid, and D. L. Weigman. 1975. Tadpoles, Ecology 69:184–192. predation, and pond habitats in the tropics. Biotropica 7:100–111. Simberloff, D., and P. Stiling. 1996. Risks of species introduced for bio- Holt, R. D. 1977. Predation, apparent competition, and the structure of logical control. Biological Conservation 78:185–192. prey communities. Theoretical Population Biology 12:237–266. Skelly, D. K. 1992. Field evidence for a cost of behavioral anti-predator Holt, R. D., and J. H. Lawton. 1994. The ecological consequences of response in a larval anuran. Ecology 73:704–708. shared natural enemies. Annual Review of Ecology and Systematics Smith, D. C. 1983. Factors controlling tadpole populations of the cho- 25:495–520. rus frog Pseudacris triseriata on Isle Royale, Michigan. Ecology Holyoak, M., and S. P. Lawler 1996. Persistence of an extinction-prone 64:501–510. predator-prey interaction through metapopulation dynamics. Ecol- Smith, D. C. 1987. Adult recruitment in chorus frogs: effects of size ogy 77:1867–1879. and date at metamorphosis. Ecology 68:344–350. Howarth, F. G. 1991. Environmental impacts of classical biological Sokal, R. R., and F. J. Rohlf. 1995. Biometry. 3rd edition. W. H. Free- control. Annual Review of Entomology 36:485–509. man, New York. Jennings, M. R., and M. P. Hayes. 1985. Pre-1900 overharvest of California Stumpel, A. H. P. 1992. Successful reproduction of introduced bull- red-legged frogs (Rana aurora draytonii): the inducement for bullfrogs Rana catesbeiana in northwestern Europe: a potential threat frog (Rana catesbeiana) introduction. Herpetologica 41:94–103. to indigenous amphibians. Biological Conservation 6:61–62. Kiesecker, J. M., and A. R. Blaustein. 1997. Population differences in Swanson, C., J. J. Cech, and R. H. Piedrahita. 1996. Mosquitofish biol- responses of red-legged frogs (Rana aurora) to introduced bull- ogy, culture, and use in mosquito control. Mosquito and Vector frogs. Ecology 78:1752–1760. Control Association of California and the University of California Kupferberg, S. J. 1995. Hydrologic and trophic factors affecting the in- Mosquito Research Program, Sacramento, California. vasion of alien bullfrogs (Rana catesbeiana) and recruitment of na- SYSTAT. 1992. Systat for Windows. 5th edition. Systat, Evanston, tive frogs (Rana boylii and Hyla regilla) in a northern California Illinois. watershed. Ph.D. dissertation. Department of Integrative Biology, Travis, J., W. H. Keen, and J. Juilianna. 1985. The role of relative body University of California, Berkeley. size in a predator-prey relationship between dragonfly naiads and Lawler, S. P. 1989. Behavioural responses to predators and predation risk larval anurans. Oikos 45:59–65. in four species of larval anurans. Animal Behaviour 38:1039–1047. Werner, E. E. 1991. Nonlethal effects of a predator on competitive in- Lawler, S. P. 1993. Direct and indirect effects in microcosm communi- teractions between two anuran larvae. Ecology 72:1709–1720. ties of protists. Oecologia 93:184–190. Werner, E. E. 1994. Ontogenetic scaling of competitive relations: size- Lawler, S. P., and P. J. Morin. 1993. Temporal overlap, competition dependent effects and responses in two anuran larvae. Ecology 75: and priority effects in larval anurans. Ecology 74:174–182. 197–213. Linden, A., and J. J. Cech Jr. 1990. Prey selection by mosquitofish Wilbur, H. M., and R. D. Semlitsch. 1990. Ecological consequences of (Gambusia affinis) in California rice fields: effect of vegetation tail injury in Rana tadpoles. Copeia 1990:18–24. and prey species. Journal of the American Mosquito Control Associ- Williamson, M. 1996. Biological invasions. Population and community ation 6:115–120. biology series. Volume 15. Chapman & Hall, London. Meffe, G. K. 1983. Effects of abiotic disturbance on coexistence of Woodward, B. D. 1983. Predator-prey interactions and breeding-pond predator-prey fish species. Ecology 65:1525–1534. use of temporary-pond species in a desert anuran community. Ecol- Menge, B. A. 1995. Indirect effects in marine rocky intertidal interac- ogy 64:1549–1555. tion webs: patterns and importance. Ecological Monographs 65: Wootton, J. T. 1994. The nature and consequences of indirect effects 21–74. in ecological communities. Annual Review of Ecology and System- Morin, P. J., S. P. Lawler, and E. A. Johnson. 1988. Competition be- atics 25:443–466.

Conservation Biology Volume 13, No. 3, June 1999